Generally, should we use Symmetrized or Conventional_Standard CIF format in VASP for Quantum Espresso calculation?

Materials Project
1

Hi,

I have been wondering about the results for calculations in VASP using symmetrized or conventional_standard cif format from materialsproject database.

For example, the compound HgBa2CuO4 (mp-6562) has 8 atoms in its conventional_standard cif:
1 Ba Ba0 0.50000 0.50000 0.69560 1.000 0.000 1a 1
2 Ba Ba1 0.50000 0.50000 0.30440 1.000 0.000 1a 1
3 Cu Cu2 0.00000 0.00000 0.50000 1.000 0.000 1a 1
4 Hg Hg3 0.00000 0.00000 0.00000 1.000 0.000 1a 1
5 O O4 0.00000 0.00000 0.79497 1.000 0.000 1a 1
6 O O5 0.00000 0.00000 0.20503 1.000 0.000 1a 1
7 O O6 0.00000 0.50000 0.50000 1.000 0.000 1a 1
8 O O7 0.50000 0.00000 0.50000 1.000 0.000 1a 1

which naturally agrees with its chemical formula.
On the other hand, the “symmetrized” format has only 5 atoms:
x y z Occ. U Site Sym.
1 Ba Ba0 0.50000 0.50000 0.30440 1.000 0.000 2h 4mm
2 Cu Cu1 0.00000 0.00000 0.50000 1.000 0.000 1b 4/mmm
3 Hg Hg2 0.00000 0.00000 0.00000 1.000 0.000 1a 4/mmm
4 O O3 0.00000 0.00000 0.20503 1.000 0.000 2g 4mm
5 O O4 0.00000 0.50000 0.50000 1.000 0.000 2e mmm .

which I think it has assumed symmetry with respect the z = 0.5 plane.
(by the way, why is this so? isn’t the lattice symmetry only given by unit (+/- 1) increments of the a,b,c lattice vectors?)

May I know which CIF type is the generally preferred choice for calculations?

I tend to use conventional_standard because of its alignment with the chemical formula.

However, in some papers reporting crystallographic positions of a material, for example,
Q. Huang et al., Phys. Rev. B 52, 462 (1995)
they tend to do so in the z=0.5 symmetrized positions.
(In this paper, they write only coordinates for 1 Ba atom only instead of 2 Ba atoms)

Hence, I wonder whether it is more common to use “conventional_standard” CIF or to use “symmetrized” cif ?
Should we expect a different result from these two?

2

Hello,
VASP does not use CIF files as calculation inputs, the program has its own format called POSCAR, which does list each atom individually. However, any program which converts CIFs to POSCARs will be able to read the symmetry operations listed at the beginning of the CIF file. For example, this material has the operation ‘-x, -y, -z’ indicating that it has at least one center of inversion symmetry, and ‘x, y, -z’ indicates the mirror symmetry across an XY plane, that you mentioned. So the 5 atomic sites which are listed in the symmetrized CIF are the minimum set of sites needed to specify the location of all the atoms; as you have guessed, the “missing” atoms are filled in by symmetry rules.

(If you’re planning to do a lot of computational material science, it is worth reading about the relations between Bravais lattices, point group symmetry, and space group symmetry)

As for which CIF you should use to generate POSCARs, I prefer the symmetrized CIF for most calculations. Enforcing the symmetry operations (and turning on ISYM = 2 in VASP if spin-orbit coupling is not considered) tends to result in a much faster calculation, but you should get the same results either way.